Chapter 6 — Calculus of Variations
Optimising over curves instead of numbers: functionals and their variations, the Euler–Lagrange equation, isoperimetric problems, the conditions that make an extremal a true minimum, and the variational methods that turn boundary value problems into minimisation problems.
In June 1696 Johann Bernoulli challenged "the sharpest mathematicians in the world" to find the curve along which a bead, sliding without friction under gravity, travels from one point to a lower one in the shortest time. The straight line is the shortest path, but not the fastest: a steeper start buys speed early. Within a year six solutions had arrived — from Johann himself, his brother Jakob, Leibniz, l'Hôpital, Tschirnhaus and, anonymously, Newton, who is said to have solved it in an evening — and all of them found the same curve, a cycloid. The problem was new in kind. Calculus finds the number $x$ at which a function is smallest; here the unknown is a whole curve, and the quantity to be minimised, the travel time, is a function of that curve. Older problems of the same kind were already famous: Queen Dido's problem of enclosing the largest area with a rope of given length, and Hero's and Fermat's principles that light takes the shortest or quickest path.
Euler (1744) and the nineteen-year-old Lagrange (1755) turned these puzzles into a calculus — the calculus of variations. Lagrange's idea was to vary the unknown curve $y$ by a small multiple of an arbitrary function $\eta$ and demand that the first-order change vanish. Out of this comes the Euler–Lagrange differential equation, which every extremising curve must satisfy and which, applied to the action $\int(T-V)\,dt$, contains all of Newtonian mechanics (Chapter 8). Legendre, Jacobi and Weierstrass then asked the harder question of when a solution of that equation is really a minimum; Riemann, Hilbert, Ritz and Galerkin reversed the logic and solved differential equations by minimising energies, which is how bridges, aircraft and microchips are analysed today by the finite element method. This chapter tells both halves of the story.
- compute first variations, prove and use the fundamental lemma and the lemma of du Bois-Reymond, and derive the Euler–Lagrange equation with all its hypotheses;
- find extremals quickly with the first integrals ($F$ free of $y$, free of $x$ — Beltrami), and solve the classical problems: shortest paths and geodesics, the brachistochrone, the minimal surface of revolution, the hanging chain, Dido's problem;
- handle several unknown functions, higher derivatives (Euler–Poisson), several independent variables (Ostrogradsky), natural boundary conditions, transversality and isoperimetric constraints with Lagrange multipliers;
- decide whether an extremal is a minimum: second variation, Legendre's condition, conjugate points and Jacobi's condition, weak versus strong minima and the Weierstrass excess function, and convexity;
- write boundary value problems and eigenvalue problems as minimisation problems, apply the Rayleigh–Ritz and Galerkin methods by hand, interpret their one-sided error bounds, and state the Dirichlet principle for Poisson's equation with its history and limits.
How this is examined
The calculus of variations is a compact, high-yield part of the CSIR NET Unit 3 syllabus. Part B questions are usually short computations: write the Euler–Lagrange equation of a given functional and evaluate the extremal at a point; recognise that a term such as $2yy'$ or $g'(x)y+g(x)y'$ is an exact derivative and drops out; use $F_{y'}=c$ or Beltrami's identity to identify a family of extremals (lines, circles, catenaries, cycloids, parabolas); impose a natural boundary condition at a free end; find the extremal of an isoperimetric problem or the minimum of $\int y'^2$ subject to $\int y^2=1$ (an eigenvalue in disguise); compute a conjugate point from the Jacobi equation; or find the coefficient of a one-term Ritz approximation. Part C, where several options may be correct, tests the logic: that the Euler–Lagrange equation is necessary but not sufficient, that extremals with given end points may fail to exist or be unique, that Legendre's condition alone does not give a minimum, that a weak minimum need not be strong, that Rayleigh–Ritz values bound eigenvalues and energies from above but pointwise values from neither side. The wrong options are almost always one of the counterexamples in the ⚠️ boxes of the two sections. GATE MA asks similar questions, and the same ideas reappear in Chapter 8 through Hamilton's principle.
Sections in this chapter
Functionals and the Euler–Lagrange Equation
Functionals and variations, the fundamental lemma, the Euler–Lagrange equation and its first integrals, the brachistochrone and the catenoid, several functions, the Euler–Poisson and Ostrogradsky equations, natural boundary conditions, transversality and isoperimetric problems.
Sufficient Conditions and Variational Methods
The second variation, Legendre's and Jacobi's conditions, conjugate points, weak and strong minima and the Weierstrass excess function; energy principles for boundary value problems, the Rayleigh–Ritz and Galerkin methods, and the Dirichlet principle.
From the CSIR NET Unit 1 course you need: differentiation under the integral sign and uniform continuity on compact sets; Taylor's theorem in two variables; the implicit function theorem and Lagrange multipliers for functions of several variables (the multiplier rule for isoperimetric problems is proved by reducing to them); the Cauchy–Schwarz and Jensen inequalities; convex functions and positive definite matrices; and the divergence theorem (Green's identities). From this course: linear second-order equations, Euler–Cauchy equations and the Sturm theory of §2.1, which solve most Euler–Lagrange and Jacobi equations; the Sturm–Liouville problems and the Rayleigh quotient of §2.3, which Chapter 6 reinterprets as minimisation problems; the Green's functions of §2.4; and Laplace's and Poisson's equations from §4.2 for the Dirichlet principle. Continuous dependence on initial data from §1.1 is used once, in the sufficiency proof of §6.2. Mechanics students will meet the same equations again in §8.1 and §8.2.